2011 1

st
semester
MIMO C i ti S t MIMO Communication Systems
#7: MIMO Receiver #7: MIMO Receiver
Kei Sakaguchi
<sakaguchi@mobile ee titech ac jp> <sakaguchi@mobile.ee.titech.ac.jp>
May 31, 2011
Schedule (2
nd
half) Schedule (2 half)
Date Text Contents
#7 May 31 A-5 MIMO receiver
#8 J 7 A 3 4 MIMO t itt #8 J une 7 A-3, 4 MIMO transmitter
#9 J une 14 B-9 Adaptive commun. system
#10 J une 21 A-6, B-14 Multi-user MIMO
#11 J une 28 B-15, 16 Distributed MIMO networks
#12 J uly 5 Standardization of MIMO #12 J uly 5 Standardization of MIMO
J uly 12 Final Examination
May 31, 2011 2 MIMO Commun. Systems (MIMO Receiver)
Agenda Agenda
Aimof today Aim of today
Derive BER & throughput performances
f b i li & li MIMO i
Contents
of basic linear & non-linear MIMO receivers
Contents
MIMO receiver architecture
F h i ti & h l ti ti Frame synchronization & channel estimation
Linear MIMO detection algorithm
N li MIMO d i l i h Non-linear MIMO detection algorithm
Measurement experiment
May 31, 2011 MIMO Commun. Systems (MIMO Receiver) 3
Warming Up Warming Up
Question 1
D ib li d t i i Describe generalized matrix inverse
of non-square matrix by using SVD. H
=
+
H
Question 2
P h h f ll i i h ld
= H
Prove that the following equation holds.
( )
H H
H H H H
1
+
=
Singular Value Decomposition
( )
m n
C

e H
m n > Non-square matrix:
SVD:
H
V U H =
m n
C

e U
m m
C

e
m m
C

e V
May 31, 2011 MIMO Commun. Systems (MIMO Receiver) 4
SVD:
V U H = C e U C e C e V
MIMO Receiver Architecture MIMO Receiver Architecture
E MMSE VBLAST QRM MLD
Frame synchronization
MIMO detection
Ex. MMSE, VBLAST, QRM-MLD
OFDM
Frame synchronization
Ch l ti ti
May 31, 2011 MIMO Commun. Systems (MIMO Receiver) 5
Channel estimation
Decision feedback
Frame Synchronization Frame Synchronization
Preamble Preamble
Preamble with repetition sequence
Training
Data
Correlation based frame detector with receive diversity
p
L
Correlation based frame detector with receive diversity

+ + +
=
1
*
p
p
) (
L
L i t i t
y y t P

=
+ + +
0
p
i
L i t i t
May 31, 2011 MIMO Commun. Systems (MIMO Receiver) 6
Cyclic Delay Preamble Cyclic Delay Preamble
Cyclic delay preamble
Tx #1
Tx #2
Tx #3
Tx #4
Without cyclic delay
With cyclic delay
Null beamforming effect
Transmit delay diversity
) (t
) (t
t
May 31, 2011 MIMO Commun. Systems (MIMO Receiver) 7
t
Experiment on MIMO Synchronization p y
Performance improvement owing to transmit & receive diversity
Performance of frame synch.
May 31, 2011 MIMO Commun. Systems (MIMO Receiver) 8
Channel Estimation Channel Estimation
1
h
At each receive antenna
T
h
1
s
Training signals
1
2
h
y
n y
T
+ = s h
2
s
n
T
+ = h s
p
S
t
M
h
t
M
s
Training signal
p p p
n h S y + =
p
p p p
| |
t p
, , ,
2 1 p
M L
M
C

e =
t
s s s S
Channel estimation error
p
L
Channel estimation
p p p p

n S h y S h
+ +
+ = =
Channel estimation error
(

=
(

=
+
2
p p
t
2
t
2
ce
E
1

E
1
n S h h
M M
o
( )
H H
p
1
p p p
S S S S

+
=


t t
M M
( ) | | ( ) | |
1
p p
2
p p
2
Tr Tr

+ +
= = S S S S
H
H
M M
o o
May 31, 2011 MIMO Commun. Systems (MIMO Receiver) 9
t t
M M
Optimal Training Sequence Optimal Training Sequence
Channel estimation error
| |
2
1
h 1
s
Training signals
( ) | |
1
p p
t
2
2
ce
Tr

= S S
H
M
o
o
1
2
h
y
2
s
p
S
Optimal training sequence
( ) | |
1
p p
Tr

S S
H
is minimal
t
M
h
t
M
s
t
t
p
p p M
H
M
PL
I S S =
( ) | |
p p
when
P L
M
p
2
t
2
ce
o
o =
p
L
Receive signal with channel estimation error
ce

N H H + = | |
M
H
2 2
t
ce ce
E
o
= N N
Effective noise
n s H n s N s H n Hs y
~

ce
+ = + + = + =
ce
| |
P L
p
ce ce
| |
r
1
~ ~
E
t
2
M
H
L
M
I n n
|
|
.
|

\
|
+ = o
May 31, 2011 MIMO Commun. Systems (MIMO Receiver) 10
p
L
|
.

\
Classification of MIMO Detector Classification of MIMO Detector
Schemes Performance Complexity
Linear
ZF
MMSE
Low Low
MMSE
VBLAST
Hybrid
VBLAST
QRM-MLD
Medium Medium
Non-linear MLD High High
May 31, 2011 MIMO Commun. Systems (MIMO Receiver) 11
Linear MIMO detection Linear MIMO detection
( (
y w w
1
st
stream
(
(
(
(

(
(
(
(

= =
r
m
y w w

1 1 11

Wy s
(

r r t t
m m m m
y w w
1
M
t
-th stream
May 31, 2011 MIMO Commun. Systems (MIMO Receiver) 12
Interference Cancellation Detection
(Zero Forcing)
R i d i l Received signal
n Hs y + =
0dB 0dB -25dB 0dB
Noise enhancement
( )
H H
H H H H W
1
+
= =
ZF weight
Noise enhancement
Estimated signal
n s n H s y H s
~

+ = + = =
+ +
n s n H s y H s + = + = =
( )
1
2
(
Effective SNR for i-th stream
( )
1
2
2
~ ~
E

=
(

H H n n
H H
o
Diversity order
( )
ZF
=
i
P

1 + = M M g
May 31, 2011 MIMO Commun. Systems (MIMO Receiver) 13
( )
1
2

ii
H
t
i
M H H o

1
t r d
+ = M M g
Minimum Mean Square Error Detection q
Cost function
| |
0dB 0dB -25dB 0dB
( ) | |
2
E Wy s W = J
MMSE weight (Wiener solution) MMSE weight (Wiener solution)
( )
0 =
c
c
W
W J
H
M
H
P
M
H I H H W
1
2
t
t

|
|
.
|

\
|
+ =
o
cW
P
. \
1
Effective SNR for i-th stream
1
1
1
2
MMSE
t

|
|
.
|

\
|
+
=

M
H
i
M
P
I H H
o

Diversity order
1
t d
+ = M M g
May 31, 2011 MIMO Commun. Systems (MIMO Receiver) 14
t
|
.

\
ii
M o
1
t r d
+ M M g
Performance of Linear Schemes Performance of Linear Schemes
No diversity gain if
t r
M M =
No performance gain of MIMO (moreover noise enhancement)
Noise
enhancement enhancement
May 31, 2011 MIMO Commun. Systems (MIMO Receiver) 15
Performance of Linear Schemes Performance of Linear Schemes
High performance due to multiplexing gain at high SNR region
Multiplexing p g
gain
May 31, 2011 MIMO Commun. Systems (MIMO Receiver) 16
Successive Interference Cancellation Successive Interference Cancellation
Diversity order increases stage by stage
times weight calculation
M
times weight calculation
t
M
Di it d
stage M M + =
t r
Diversity order
May 31, 2011 MIMO Commun. Systems (MIMO Receiver) 17
Performance of Hybrid Scheme Performance of Hybrid Scheme
Diversity gain owing to successive interference cancellation
Diversity gain
May 31, 2011 MIMO Commun. Systems (MIMO Receiver) 18
Performance of Hybrid Scheme Performance of Hybrid Scheme
Diversity gain improves the throughput performance at lower SNR region
Diversity gain Diversity gain
May 31, 2011 MIMO Commun. Systems (MIMO Receiver) 19
MaximumLikelihood Detection Maximum Likelihood Detection
2
Maximum likelihood detection
2
~
~
min arg

s H y s
s
=
2 Tx with QPSK signaling
Minimumdistance
Pair wise error probability
Receive signal
Minimum distance
(
(

=
2
2
2
) (
erfc
2
1
) (
o
k i
k i
p
s s H
s s
h
Receive signal
(

2 2 o
r
2
1
M
ik
d P

|
|
|

|
+ s
1
h
2
h
Diversity order
M g =
t
2
4
1
M
|
.

\
+ s
o
Computational cost
t
M
M
r d
M g =
May 31, 2011 MIMO Commun. Systems (MIMO Receiver) 20
t
ary
M
M
combinations on
r
M receiver branches
Performance of Non-Linear Scheme Performance of Non Linear Scheme
Full receive diversity due to MLD at all receive branches
Full receive diversity
May 31, 2011 MIMO Commun. Systems (MIMO Receiver) 21
Performance of Non-Linear Scheme Performance of Non Linear Scheme
Maximal available throughput among all MIMO detection schemes
It t ti l l it i l t hibit d Its computational complexity is almost prohibited
Full receive diversity
May 31, 2011 MIMO Commun. Systems (MIMO Receiver) 22
QRM-MLD QRM MLD
QRM-MLD approaches the performance of MLD asymptotically
QRM-MLD decreases the computational complexity drastically QRM MLD decreases the computational complexity drastically
(
(
(

(
(
(

=
t
m
s r r

1 1 11
0 Rs
(
(

(
(

=
t t t
m m m
s r

0 0
0 Rs
QR H U t i l d iti
May 31, 2011 MIMO Commun. Systems (MIMO Receiver) 23
QR H = Upper triangular decomposition:
Performance of Hybrid Scheme Performance of Hybrid Scheme
QRM-MLD achieves full receive diversity as with MLD
There is SNR penalty due to error propagation There is SNR penalty due to error propagation
SNR penalty
May 31, 2011 MIMO Commun. Systems (MIMO Receiver) 24
Performance of Hybrid Scheme Performance of Hybrid Scheme
In IID, performance of VBLAST & QRM-MLD are almost the same
SNR penalty
May 31, 2011 MIMO Commun. Systems (MIMO Receiver) 25
Comparison of MIMO Detection Schemes p
Multiplexing gain Diversity gain p g g y g
MMSE
VBLAST stage M M +
t r
1
t r
+ M M
t
M
t
M
MLD
QRM-MLD
r
M s
g
t r t
t
M
t
M
r
M
Complexity
r
t
MMSE
VBLAST
) ( (INV)
r t
M M O L O +
) ( 2 (INV)
r t t
M M O L O M +
M
MLD
QRM-MLD
) 2 / ) 1 ( ( ) QRD (
t t r t
+ + M MM M M O L O
) (
t
ary r t
M
M M M O
May 31, 2011 MIMO Commun. Systems (MIMO Receiver) 26
: L
Number of symbols
Measurement Experiment Measurement Experiment
Measurement condition
MIMO structure 4 (Tx) x 4 (Rx)
Array structure Half a wavelength spacing Array structure Half a wavelength spacing
linear array
MIMO transmit scheme Spatial multiplexing MIMO transmit scheme Spatial multiplexing
MIMO receive scheme ZF, MMSE, VBLAST, MLD
Modulation BPSK QPSK 16QAM 64QAM Modulation BPSK, QPSK, 16QAM, 64QAM
Frame structure IEEE802.11a based
(without coding) (without coding)
Packet length 480 bits
Measurement points 40cmx 40cm(2cmstep)
May 31, 2011 MIMO Commun. Systems (MIMO Receiver) 27
Measurement points 40cm x 40cm (2cm step)
Measurement Environment Measurement Environment
May 31, 2011 MIMO Commun. Systems (MIMO Receiver) 28
Propagation Characteristics Propagation Characteristics
Randompathloss distribution in space domain Random pathloss distribution in space domain
Slight spatial correlation will affect MIMO performance
Pathloss distribution Spatial correlation
May 31, 2011 MIMO Commun. Systems (MIMO Receiver) 29
SISO Transmission Performance SISO Transmission Performance
BER performance agrees with that in Rayleigh environment
Degradation of throughput performance is due to
imperfection of frequency characteristics in RF transceiver
BER performance
Throughput performance
May 31, 2011 MIMO Commun. Systems (MIMO Receiver) 30
MIMO Transmission Performance MIMO Transmission Performance
Performance degradation of BER is due to spatial correlation g p
RF impairment & spatial correlation degrades throughput performance
BER performance
Throughput performance
May 31, 2011 MIMO Commun. Systems (MIMO Receiver) 31
Summary Summary
MIMO receiver architecture
T it & i di it f h i ti Transmit & receive diversity on frame synchronization
Orthogonal preamble sequences for MISO channel estimation
MIMO detection algorithmto resolve the transmit signal MIMO detection algorithm to resolve the transmit signal
MIMO detection algorithm
Multiplexing gain to improve throughput performance p g g p g p p
Diversity gain to improve BER performance
Tradeoff between complexity & performance (diversity order)
Li h (ZF MMSE) Linear schemes (ZF, MMSE)
Non-linear schemes (VBLAST, MLD, QRM-MLD)
Further improvement by transmit MIMO schems
May 31, 2011 MIMO Commun. Systems (MIMO Receiver) 32
Further improvement by transmit MIMO schems